[address-policy-wg] RE: The price of address space

Jeroen Massar wrote:

>>NAT, in general including CGN, does scale to the extent to make
>>IPv6 not necessary.

> It scales on paper, till you start using it.
> 
> 1 IPv4 address, 65k TCP ports, if one user opens maps.google, he uses
> 200 TCP sessions on average, thus 65k/200 == 332 users per IPv4 address.

200 is a peak of an extra ordinary application and, even with the
application, unused ports are closed quickly that dynamic allocation
of ports save average port requirements.

Note also that consumption of extra ordinary number of port is a
state maintenance problem for NAT between 6 and 4.

> Yes, indeed, really "scales" well.

Yes, of course.

> CGNs will btw only delay the inevitable.

It is inevitable that protocols have lifetime, of course.

A problem is that the lifetime of IPv6 has already expired.

> (On the subject of CGN and content-restricting ISPs for CP and other

Let's not discuss it here, because there are alternative ways to NAT.
 
>>If it is 10 years, we should use not IPv6 but something else.

> You can do that, your competition will love you for it.

Considering that route aggregation is also an important goal of
address policy, which IPv6 failed to address, we do need something
else.

>>As you don't accept the answer "never", discussion on whether it will
>>actually be "never" or not is inevitable.

> Give me one valid technical reason why I would accept "never"?

Because, as you failed to argue against, it is no better than IPv4
with NAT.

						Masataka Ohta